Assessing Long Distance Communication Alternatives for the Remote Control of AGVs

Abstract

Remote monitoring and control of factory equipment promises a more streamlined and therefore less expensive system operation and maintenance. The geographical distance between a factory and its control center, however, may influence the Quality of Service parameters of the network connections which might stymie the overall control process. To get a better understanding of these potential issues and their impact, we conducted a series of measurements over varying distances for the remote control, operation and simulation of Automated Guided Vehicles (AGVs) that are often used in modern factory environments. To achieve these tests, we defined three communication patterns reflecting local and remote connections as well as the usage of cloud-based services. Applying these patterns, we connected the Factory of the Future at the Aalto University in Finland with the VxLab at the RMIT University in Australia and the Microsoft Azure cloud in the Netherlands. This allowed us to measure important Quality of Service networking parameters for the communication over short, medium, and very long distances. In this paper, we present first empirical results and discuss their impact on the remote control of AGVs.

Simulation Components in Gazebo

“…The MirrorLabs project features the development of MR applications primarily for educational purposes [9]. Our work continues ideas sketched in [10] where we evaluated latency issues in distributed Gazebo usage scenarios. Related to our Gazebo work, production line planning using VR has been studied in [11] and some work on Functional Mockup Interface-based simulation is presented in [12].…”

Towards enhanced live visualization based on communication delay prediction for remote AGV operation

“…In the future, the visualization server The main difference between the simulated and the live stream localization data is that the latter travels a long distance from the robot to the to the visualization server with variable delay (jitter). The benchmark for such delay is approximately two to three times the expected RTT [8], calculated from the geographical separation over the speed of light through fiber optic [11]. Nevertheless, especially while using an internetbased channel, the delay behavior is subject to many external factors.…”

“…2. The deployment is based in shop-floor components interconnected to a visualization cloud server through a point to site VPN tunnel, in such way it is possible to have seamless communication through a secure channel over Internet Protocol v4 (IPv4), according the communication patterns from [8]. As any other web application, the cloud server deployment is expected to be as near as possible to the clients.…”

“…A sampling window size (S s ) of 4, trained for 20 epochs provided the best loss (2.7 * 10 −4 ) and Root Mean Square Error (2.8). The data-set used for training and test consists of 259200 registers of timestamped RTT captured during 3 days at the rate of 1 sample per second, following the cloud communication pattern from [8]. While this component is relevant for the DP algorithm, this article focuses on the procedural context.…”

“…Ohori et al exalted the importance of handling correctly the wireless communication inherent in AGVs [7]. The authors 978-1-7281-2989-1/21/$31.00 ©2021 IEEE presented a survey about delay and throughput for longdistance communication of AGVs in [8].…”